Predictive Maintenance in IoT Environments Using Machine Learning: Opportunities, Challenges, and Innovations

Authors: Simran Sethi

DOI: https://doi.org/10.5281/zenodo.14945146

Short DOI: https://doi.org/g86n7k

Country: USA

Full-text Research PDF File:   View   |   Download

Abstract: Predictive maintenance (PdM) is arguably one of the best and most innovative uses of the Internet of Things (IoT) and machine learning (ML). Using a vast amount of sensor data, PdM systems aim at predicting and preventing equipment failures by scheduling interventions timely so that the machinery does not reach an undiagnosed state of inoperability. This document surveys the most relevant aspects of IoT in PdM concentrating on IoT’s machine learning models and deep learning frameworks for real time analytics. We outline an industrial high level PdM architecture for IIoT based on deep learning and anomaly detection with reinforcement learning. Furthermore, we elaborate on notable real factory deployment issues such as data silos, scalability of defined models, model interpretability, human factors, and benchmark problems. Lastly, we point out the remaining gaps that aim to extend PdM from proof-of-concept models to widespread industrial use.

Keywords: Predictive Maintenance (Pdm), Industrial Internet Of Things (Iiot), Machine Learning, Deep Learning, Reinforcement Learning, Anomaly Detection, Remaining Useful Life (RUL), Edge-Cloud Architecture, Sensor Data, Concept Drift, Model Interpretability, Real-Time Analytics

Paper Id: 232173

Published On: 2023-08-09

Published In: Volume 11, Issue 4, July-August 2023

Share this